(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
[11:16] Andrew.Hellershanks @tanglegrid.net:8002: It is about quarter past. Let's get started.
[11:16] Vincent Sylvester: I have local chat logging turned on in the viewer so I'll just grab it from there
[11:16] Ubit.Test @cc.opensimulator.org:8002: ty
[11:19] Andrew.Hellershanks @tanglegrid.net:8002: The big(?) news this week is the bump in the version number of Open Simulator. It is now 0.9.3.1Dev.
[11:19] Ubit.Test @cc.opensimulator.org:8002: errr guess the release of 0.9.3.0 is a little bigger news :p
[11:20] Andrew.Hellershanks @tanglegrid.net:8002 nods to Ubit.
[11:20] Andrew.Hellershanks @tanglegrid.net:8002: Nice to have a new version in time for the OSCC.
[11:20] Ubit.Test @cc.opensimulator.org:8002: why do you think ... nm :p
[11:20] Lyr.Lobo @cc.opensimulator.org:8002: OSCC is upgraded *grins*
[11:21] Lyr.Lobo @cc.opensimulator.org:8002: and we are getting ready for speaker training on Saturday at 9 am Pacific
[11:21] Ubit.Test @cc.opensimulator.org:8002: well code and bug reports was quiet and it was more than time to release
[11:21] Lyr.Lobo @cc.opensimulator.org:8002 nods
[11:21] Lyr.Lobo @cc.opensimulator.org:8002: Thank you, Ubit
[11:22] Andrew.Hellershanks @tanglegrid.net:8002: Now we will have at least one thing to talk about during the morning keynote panel. :)
[11:22] Lyr.Lobo @cc.opensimulator.org:8002 grins
[11:22] Ubit.Test @cc.opensimulator.org:8002: yeah we did release a version with the things you told you about last year :P
[11:22] Ubit.Test @cc.opensimulator.org:8002: we...
[11:23] Vincent Sylvester: You did that, the royal we :)
[11:23] Ubit.Test @cc.opensimulator.org:8002: oh someone did look the the site!!
[11:23] Ubit.Test @cc.opensimulator.org:8002: :)
[11:23] Vincent Sylvester: I ran upgrades here two weeks ago, went mostly smoothly after doing some manual stuff
[11:24] Lyr.Lobo @cc.opensimulator.org:8002: Great
[11:24] Vincent Sylvester: I deployed the new migration system I been working on here, which gave me some useful information on what I had missed to change
[11:24] Vincent Sylvester: I been actively in debugging phase with that now, looking promising
[11:25] Vincent Sylvester: Couple more weeks in production testing and it should be ready for Ubit to pick apart :)
[11:25] Ubit.Test @cc.opensimulator.org:8002: (glupss)
[11:25] Lyr.Lobo @cc.opensimulator.org:8002 chuckles
[11:26] Andrew.Hellershanks @tanglegrid.net:8002: :)
[11:26] Vincent Sylvester: So far dotnet version has shown an overall reduction in cpu usage by as much as 50% which is quite big
[11:26] Ubit.Test @cc.opensimulator.org:8002: not sure if osg will be back for next meeting
[11:27] Lyr.Lobo @cc.opensimulator.org:8002: I can tweet this meeting location if you wish
[11:27] Ubit.Test @cc.opensimulator.org:8002: guess those not here also don't look to X :)
[11:28] Vincent Sylvester: You mean twitter ;)
[11:28] Ubit.Test @cc.opensimulator.org:8002: not the first time here, i remember those cookies from our last meeting here
[11:29] Vincent Sylvester: Rest assured they are the same ones xD
[11:29] Ubit.Test @cc.opensimulator.org:8002: :)
[11:30] Ubit.Test @cc.opensimulator.org:8002: well motoko we did already covered most opensim news of last week :)
[11:30] Ubit.Test @cc.opensimulator.org:8002: just that 0.9.3.0 is now release
[11:32] Vincent Sylvester: Oh reminds me, speaking of cpu, it seems that has drastically changed, because the function I used to get the cpu usage now doesn't function anymore
[11:32] Vincent Sylvester: The particle tube behind me should show few particles when cpu usage is low, but the reported number is now way off
[11:33] Vincent Sylvester: Haven't found a clear solution for that yet either
[11:33] Lyr.Lobo @cc.opensimulator.org:8002 nods
[11:33] Vincent Sylvester: Very often just reports 0 usage even when probing multiple times over a couple seconds
[11:33] Motoko.Karu @login.digiworldz.com:8002: 0_o
[11:33] Ubit.Test @cc.opensimulator.org:8002: there is no direct function to get cpu usage
[11:34] Lyr.Lobo @cc.opensimulator.org:8002: Yes, a Tweet is a Twitter message over X
[11:34] Ubit.Test @cc.opensimulator.org:8002: :)
[11:34] Vincent Sylvester: No that's true. I had compared processor tick time for that, which seemed accurate enough under mono. dotnet reports like a tenth now
[11:35] Vincent Sylvester: I tried longer wait times, but no dice. Will have to dig into that
[11:35] Vincent Sylvester: I found one thread talking about this saying that the processor or thread count now has to be accounted for, which is a lot more work
[11:36] Vincent Sylvester: On windows you have diagnostics that will tell you directly, but that's obviously not platform independent
[11:36] Ubit.Test @cc.opensimulator.org:8002: yeah
[11:37] Vincent Sylvester: Very hard to work out if that works when you have to artificially create cpu usage to get readings and your cpu is just too powerful to register any percentage
[11:38] Ubit.Test @cc.opensimulator.org:8002: :)
[11:38] Vincent Sylvester: When you want to stress test a region it somehow is much harder than accidentally finding something really heavy, guess I have to try raycasts or something
[11:38] Ubit.Test @cc.opensimulator.org:8002: map gen :p
[11:39] Vincent Sylvester: Oh yeah I forgot about that, true that might work well
[11:39] Ubit.Test @cc.opensimulator.org:8002: but its single thread
[11:39] Ubit.Test @cc.opensimulator.org:8002: hmm mostly..
[11:40] Ubit.Test @cc.opensimulator.org:8002: ( and dy design )
[11:40] Ubit.Test @cc.opensimulator.org:8002: by...
[11:40] Ubit.Test @cc.opensimulator.org:8002: it is supposed to be a low priority task
[11:41] Vincent Sylvester: Yeah it sticks to 100% cpu if you put it into its own thread. There might be some way to parallel some tasks in there, but I have dug much into Warp3D yet
[11:41] Ubit.Test @cc.opensimulator.org:8002: you do not want to do that
[11:41] Ubit.Test @cc.opensimulator.org:8002: regiosn have better use for the cpu
[11:42] Vincent Sylvester: Yeah, but can always look at limiting usage and still reduce the time it takes to make a tile, more efficiency than just raw power
[11:44] Vincent Sylvester: I think there is potential there, just finding it might be quite difficult, lots of abstract stuff and trig math :)
[11:45] Vincent Sylvester: Probably start with working out a way to profile it to see which part takes the longest and go from there
[11:46] Vincent Sylvester: Given the lower cpu usage of dotnet in general I would think it might already have improved how fast it can generate a tile as well
[11:46] Vincent Sylvester: Heaviest region I tested with took about 10 minutes for a tile, that's not that bad considering you only really do that maybe once a day
[11:48] Vincent Sylvester: There is some caching built into it, but I wonder if caching the results of object rendering and only rendering the changed objects might not ultimately improve performance, at least at runtime
[11:52] Andrew.Hellershanks @tanglegrid.net:8002: I've been meaning to run my own load test using a bad script I picked up some time ago.
[11:54] Andrew.Hellershanks @tanglegrid.net:8002: It is a script for blowing bubble gum. Under the 0.8 it had a high runtime. I want to see how its runtime compares to the numbers we get from 0.9
[11:55] Vincent Sylvester: So far what I have found is that scripts feel more responsive probably due to faster work through the queue of actions they request to perform, so you have less "stack" to go through before you get a result
[11:55] Vincent Sylvester: Whether that is measurable though, probably have to have something very heavy for that yeah
[11:56] Ubit.Test @cc.opensimulator.org:8002: not only dotnet does a few things faster and so does our code.. here or there
[11:56] Andrew.Hellershanks @tanglegrid.net:8002: Changes made to OS since the 0.8 days may have changed how much of a load it imposes on a system.
[11:56] Ubit.Test @cc.opensimulator.org:8002: may ? lol
[11:57] Vincent Sylvester: I would hope they did given all the work that went into just that avx thing
[11:57] Andrew.Hellershanks @tanglegrid.net:8002: Different script engine, dotnet vs mono. Yeah just a small change or two. :)
[11:58] Ubit.Test @cc.opensimulator.org:8002: sorry to tell Andrew.Hellershanks but 0.8x was not the best opensim version...
[11:59] Ubit.Test @cc.opensimulator.org:8002: by far...
[11:59] Andrew.Hellershanks @tanglegrid.net:8002: It probably was the best until something newer came along. :)
[11:59] Ubit.Test @cc.opensimulator.org:8002: hmm that needs testing :P
[12:00] Andrew.Hellershanks @tanglegrid.net:8002: We are at the top of the hour. If there isn't anything else to discuss we can conclude the meeting now, or we can go another 15 minutes as we waited for other regulars to find us here.
[12:01] Vincent Sylvester: One thing I am curious about is how it handles overload. Mono would regularly just grind to a halt completely and even after cutting the load not recover. Scheduling difference may make this possible now, not that it is something that should happen ideally
[12:02] Vincent Sylvester: Some SL scripts are notorious for that, few of them and boom region is deadlocked
[12:03] Vincent Sylvester: Might give some room now to get the script tickrate in sync with SL
[12:03] Vincent Sylvester: Not that I think it's particularly necessary in the first place
[12:03] Ubit.Test @cc.opensimulator.org:8002: scripts have no tickrate
[12:03] Vincent Sylvester: They kinda do
[12:04] Vincent Sylvester: Not as in the definition of a tickrate, but the maximum commands they'll run each frame? something like that
[12:04] Ubit.Test @cc.opensimulator.org:8002: not at opensim
[12:05] Vincent Sylvester: At SL it's 15, we get 10, but you can change that in config to be higher
[12:05] Vincent Sylvester: Shouldn't ever hit that though, that's bad script design
[12:06] Ubit.Test @cc.opensimulator.org:8002: sl was a single thread engine doing bits of scripts on main heartbeat... diferent thing
[12:07] Ubit.Test @cc.opensimulator.org:8002: original opensim was more like fire 300 threads and pray
[12:08] Ubit.Test @cc.opensimulator.org:8002: ofc half of those gtot stuck on llSleeps :p
[12:10] Lyr.Lobo @cc.opensimulator.org:8002 smiles
[12:10] Vincent Sylvester: If you run a fast timer and iterate a count divided by llGetTime() you get a "tickrate", which will be set to the minimal timer interval, but you can run the same division in a while loop to get the ultimate rate at which it can perform actions. Yes that is silly, I am well aware
[12:11] Ubit.Test @cc.opensimulator.org:8002: timers to have a min tick... diferent story
[12:11] Ubit.Test @cc.opensimulator.org:8002: do have..
[12:11] Vincent Sylvester: It's something to do with the async udp commands if I recall
[12:11] Vincent Sylvester: Config setting, not to be messed with I must add
[12:11] Vincent Sylvester: Scripts that hit that limitation are just badly designed
[12:12] Andrew.Hellershanks @tanglegrid.net:8002: There are plenty of badly designed scripts floating around.
[12:12] Lyr.Lobo @cc.opensimulator.org:8002 whispers: I'll keep an eye on the meeting log page to see if we are meeting again here next week. http://opensimulator.org/wiki/Office_Hours I must dash, dear friends, and thank you for hosting the meeting, Ubit, Andrew, and Vincent for the lovely location. Motoko, always great to see you. Have a grand week!
[12:12] Lyr.Lobo @cc.opensimulator.org:8002: not a whisper *grins*
[12:12] Lyr.Lobo @cc.opensimulator.org:8002: I'll keep an eye on the meeting log page to see if we are meeting again here next week. http://opensimulator.org/wiki/Office_Hours I must dash, dear friends, and thank you for hosting the meeting, Ubit, Andrew, and Vincent for the lovely location. Motoko, always great to see you. Have a grand week!
[12:12] Ubit.Test @cc.opensimulator.org:8002: cya Lyr :)
[12:12] Andrew.Hellershanks @tanglegrid.net:8002: tc, Lyr. See you next week.
[12:13] Lyr.Lobo @cc.opensimulator.org:8002: Have a great day!
[12:13] Ubit.Test @cc.opensimulator.org:8002: well guess we also only had one thing to tell :), that was even old news :)
[12:14] Andrew.Hellershanks @tanglegrid.net:8002: yup
[12:14] Lyr.Lobo @cc.opensimulator.org:8002: Many thanks!
[12:15] Andrew.Hellershanks @tanglegrid.net:8002: It is now officially an hour since we started. Any last minute items? If not, I will call this meeting to a close.
[12:16] Andrew.Hellershanks @tanglegrid.net:8002: I wish the best of luck to the admins at OS Grid that they will get their maintenance task done quickly and without further hiccups. If not, we may be back here again next week.
[12:16] Andrew.Hellershanks @tanglegrid.net:8002: Thank you all for coming. See you again next week.
